In order to facilitate the utilization of healthcare data for research purposes while ensuring patient privacy, the removal of personal health information (PHI) is imperative. This process, commonly referred to as de-identification or anonymization, involves identifying and obfuscating PHI within clinical text. Although open source systems capable of identifying such information have been developed primarily for the English language, recent progress has been made in the field of Spanish language de-identification. However, there remains a dearth of open source models specifically tailored for accurate identification of PHI in Spanish clinical text. This study aims to address this gap by evaluating state-of-the-art open source deep learning models for their efficacy in recognizing a subset of PHI elements. To accomplish this, a comprehensive assessment of all pre-trained models available in the Hugging Face model hub was conducted to ascertain their suitability for PHI identification in Spanish clinical texts. Furthermore, an automated framework was developed to facilitate the evaluation of these open source deep learning models in the context of PHI recognition within Spanish clinical text. These models are evaluated with the aim of identifying the best ones, providing a foundation for subsequent implementation in various fields of research, specifically in the context of anonymization. To evaluate the performance of the models, the MEDDOCAN corpus, comprising 500 annotated Spanish clinical notes, was utilized. The results of the analysis revealed four models that exhibited the highest performance. These models demonstrated favorable metrics in terms of precision, recall, and F1 score. Moreover, it was observed that leveraging the entirety of the text and maximizing the number of tokens provided to the models resulted in enhanced performance, as demonstrated in this study.